For over a decade Deft Research has studied the behavior of the health insurance consumer across all lines of insurance business. Combining this knowledge with “big data” analytics, we have created a powerful set of models that allow clients to predict the relative likelihood that an individual prospect or member will exhibit a key characteristic or behavior. Our predictive models help health insurers develop precise, targeted campaigns aimed at certain segments of prospects or members.

Deft’s Predictive Models allow marketers to:

  • Develop precise, targeted campaigns.
  • Align the right product offers to the right consumers.
  • Convert prospects into customers.
  • Tailor the most effective message for a specific prospect.

  List Scoring Service

Marketing health insurance is challenging. Mailing lists may contain many households who are not eligible to purchase coverage or that have a strong preference for products you don’t offer. Online shoppers often lack the discipline to sift through troves of insurance options to find what you provide. And, your call center may be overwhelmed assisting prospects who would be more than willing to complete an application online.

Deft Research’s List Scoring Service can help prioritize your sales, marketing and outreach efforts across multiple channels.

Our Prospect Models include:

  • Insurance Source Models
  • Product Preference Models
  • Shopping and Switching Models
  • Direct Response Models

  Retention Modeling

Once a prospect becomes a member you have to keep them satisfied. Member churn is expensive and dissatisfied members can dampen quality ratings and possibly reimbursement.

Deft Research’s Member Models help identify which members are likely to leave, shop, or be dissatisfied. They can also help identify the most effective communications channels for each member.

Our Member Models include:

  • Retention Models
  • Satisfaction Models
  • Communication Models

  How do the models work?

Deft Research’s predictive models incorporate the following data:

  • Consumer demographic data (either client-supplied or third party*)
  • Zip Code level household characteristics
  • County level local market competitive characteristics

*If not supplied by client, Deft Research will obtain third party marketing data for an additional charge.

Using these data elements, our standard models create a propensity score for each modeled behavior/attribute that can be applied to each prospect or member. These scores can be added to a prospect mailing list or can be applied in real time during an online shopping session or a customer service call. Clients use these scores to maximize conversion rates, improve program compliance, and to reduce churn.


Our models are designed to be “layered”. This means that results can be improved by applying multiple models or by adding our models to your existing models. Here’s an example:

A health insurer and their agency are designing a campaign to sell individual Medicare Supplement policies into a new geographic market using a combination of direct mail and seminars. The client has purchased a list of all senior households in the target geographic market. The particular market has a high incidence of consumers covered by employer-sponsored retiree plans and Medicaid. Here is an example how the Deft Research’s predictive models can be used to maximize response to the campaign:

  1. Using the Insurance Source Models, the households with consumers most likely covered by Medicaid and employer-sponsored retiree plans are flagged and removed from the list.
  2. Using the Product Preference Models, the households most likely to have a preference for Medicare Supplement (vs. Medicare Advantage) are identified and flagged.  Households with low preference scores for Medicare Supplement are removed from campaign.
  3. Using the Shopping and Switching Models, individuals who are likely to shop or switch coverage are identified. 
  4. The client’s agency has a well-established direct mail response model.  Deft Research’s Direct Response Models indicate a prospect's relative likelihood to respond to direct mail or phone and is used in combination with the agency model to determine if a combination approach improves results.

By applying this “layered” approach, response rates and conversions are greatly improved versus the application of a direct response model alone.